Multiple Semantic Web Languages (Re: today 16 March)

From: Sandro Hawke (
Date: 03/16/04

  • Next message: Benjamin Grosof: "regrets Re: Joint Committee telecon today 16 March"
    > Multiple Semantic Web Languages [3] (if Sandro is available) (all - 20 min)
    [ I expect to attend.  The fetus is still awaiting a Director's
    Decision or something, I don't know.  I'll have to switch to my mobile
    phone around 5pm, though, if we run late as usual. ]
    There was a related thread in WebOnt last week.  The archives are
    publically accessible:
    I observed that aside from the politics and perception, there are some
    decent reasons to carrying other language expressions in RDF:
          - You'll have an RDF/XML parser and triplestore anyway
          - RDF/XML users should have a smooth migration path
          - When ontologies are just more RDF data, RDF systems
            for handling provenance, pub/sub, trust, etc can be
            directly applied
    But these might be served by putting SWRL expressions inside RDF
    literals, too.   The syntactic similarity between terms and clauses in
    eg Prolog can be rather useful, while the semantic differences can be
    quite confusing.  Compare/contrast Java with Scheme.  Which is more
    elegant?  Which is more successful?  Which boat do you want to be in?
    This may be a hard point around which to find consensus.
    My sense is still:
         - SWRL should have a presentation syntax (something like Prolog
           or InfixKIF, as Harold was working on, or N3); I believe we
           agreed on this in principle last time.  That syntax should be
           seen as a first-class exchange protocol, (unlike n-triples and
           InfixKIF, which were disclaimed as second-class special purpose
           languages by their creators.)  Since SWRL is an extension of
           (at least a subset of) RDF, this serves also as an alternative
           serialization for (at least some) RDF graphs.
         - SWRL should not have an "RDF-Like" syntax, since that would
           cause RDF systems to do odd and problematic things with SWRL
           axioms.  SWRL should either have an RDF syntax or nothing like an
           RDF syntax. 
     -- sandro

    This archive was generated by hypermail 2.1.4 : 03/16/04 EST